Short term online corrosion measurements in biomass fired boilers. Part 2: Investigation of the corrosion behavior of three selected superheater steels for two biomass fuels

Fuel Processing Technology 142 (2016) 59–70

Contents lists available at ScienceDirect

Fuel Processing Technology journal homepage: www.elsevier.com/locate/fuproc

Short term online corrosion measurements in biomass fired boilers. Part 2: Investigation of the corrosion behavior of three selected superheater steels for two biomass fuels Stefan Retschitzegger a,⁎, Thomas Gruber a, Thomas Brunner a,b,c, Ingwald Obernberger b,c a b c

BIOENERGY 2020+ GmbH, Inffeldgasse 21b, 8010 Graz, Austria Institute for Process and Particle Engineering, Graz University of Technology, Inffeldgasse 13, 8010 Graz, Austria BIOS BIOENERGIESYSTEME GmbH, Inffeldgasse 21b, 8010 Graz, Austria

a r t i c l e

i n f o

Article history: Received 3 April 2015 Received in revised form 15 September 2015 Accepted 17 September 2015 Available online xxxx Keywords: Biomass combustion High-temperature corrosion Online corrosion measurements

a b s t r a c t The high temperature corrosion behavior of the boiler steels 13CrMo4-5 (1.7335), P91 (1.4903) and 1.4541 has been investigated during short-term test runs (~500 h) at a biomass fired grate furnace combined with a drop tube. For the test runs performed with 13CrMo4-5 and P91 chemically untreated wood chips have been used as fuel, whereas waste wood has been used for test runs with P91 and 1.4541. Online corrosion probes and a mass loss probe have been used applying a methodology developed in a previous study to correct for a measurement error occurring during short-term measurements with online corrosion probes (mass loss correction). Furthermore, deposit probe measurements have been performed to evaluate the deposit build-up rate and the chemical composition of deposits. SEM/EDX analyses of the corrosion probes have been performed subsequently to the test runs to gain information regarding the chemical composition and structure of the deposits as well as the corrosion layers. The furnace has been operated at constant load to ensure constant combustion conditions. The flue gas temperature at the probes has been varied between 740 and 900 °C and the probe surface temperature has been varied between 400 and 560 °C in order to determine their influence on the corrosion rate. General trends determined by the variation of these temperatures were similar for all boiler steels: the corrosion rate increased with increasing flue gas temperature and also with increasing probe surface temperature. For chemically untreated wood chips combustion at low flue gas temperatures (740 °C) the corrosion rates were comparable for 13CrMo4-5 and P91 at all probe surface temperatures. However, at flue gas temperatures of 800 °C and higher P91 showed better corrosion resistance than 13CrMo4-5. For waste wood combustion 1.4541 generally showed a better corrosion resistance than P91. The mass loss correction of the measurement error occurring in the initial phase resulted in different errors of 55% for 13CrMo4-5 and 32% for P91 for chemically untreated wood chips. For waste wood the mass loss correction resulted in errors of 55% for P91 and 77% for 1.4541. The results from the mass loss determination for the waste wood test runs scattered stronger compared to the wood chips test runs. Therefore, the fits were not that accurate and the error margin was higher. However, the results outline that the mass loss correction is relevant in order to achieve a meaningful comparison of different short-term test runs using online corrosion probes. © 2015 Elsevier B.V. All rights reserved.

1. Introduction and objectives In a previous study [1] the application of online corrosion probes for the evaluation of high temperature corrosion in biomass-fired boilers based on short-term measurements has been investigated. It was shown that during the initial phase the measurement does not represent the actual corrosion rate, but the corrosion signal starts at zero and increases gradually until it is proportional to the corrosion rate.

⁎ Corresponding author.

http://dx.doi.org/10.1016/j.fuproc.2015.09.021 0378-3820/© 2015 Elsevier B.V. All rights reserved.

This “start-up” effect results in an overestimation of corrosion rates. By the application of an additional mass loss probe a methodology was developed which allows for the correction of this error and results in a significant improvement of the accuracy of short-term corrosion probe measurements. Several studies on high temperature corrosion of boiler steels report significant differences in the compositions and the structures of corrosion layers depending on the composition of the steel [2–7]. Corrosion layers of low-alloyed steels such as 13CrMo4-5 mainly consist of iron oxides [5,6] whereas for steels containing higher shares of chromium separate layers of iron oxides and chromium oxides were reported [7].

60

S. Retschitzegger et al. / Fuel Processing Technology 142 (2016) 59–70

These differences may also influence the “start-up” effect during online corrosion probe measurements. Therefore, the objective of the work presented was to demonstrate the performance of short-term test runs with online corrosion probes and mass loss probes using three boiler steels as well as an evaluation of the results based on the new methodology and a comparison between the different materials. Within the work presented the steels 13CrMo4-5, P91 and 1.4541 have been evaluated. The test runs have been performed using a 50 kW grate furnace combined with an electrically heated drop tube. For this purpose chemically untreated wood chips and waste wood have been used as fuels. Subsequently to the test runs, the corroded steels have been analyzed to determine the corrosion mechanisms prevailing.

2. Materials and methods The test runs have been carried out in a biomass grate furnace combined with a drop tube using chemically untreated forest wood chips and waste wood as fuels. The chemically untreated forest wood chips fuel mainly consisted of logging residues but also contained bark as well as small fractions of fine particles and needles. The biomass originated from a local supplier close to Graz, Austria and was harvested within a radius of approximately 50 km around Graz. For the two test runs, two different batches of fuels were supplied. In the following, this fuel is referred to as “wood chips”. The waste wood was provided by a local supplier in the area of Graz and mainly consisted of demolition wood. For these test runs only one batch of fuel was used. The fuels used complied with the specifications according to EN 14961-1:2010 [8] as presented in Table 1. The chemical compositions of the fuels can be found in Section 3.1, Table 4. Within the tests, the corrosion behavior of the superheater materials 13CrMo4-5, P91 and 1.4541 was investigated. The chemical compositions of the steels are shown in Table 2. For the combustion of chemically untreated wood chips usually low and medium alloyed steels are applied for superheaters, therefore only 13CrMo4-5 and P91 have been investigated for this fuel. In case of waste wood combustion a higher risk for corrosion is assumed and therefore only the two higher alloyed steels P91 and 1.4541 have been investigated. The materials have been chosen, since these are materials which are typically used for superheaters in biomass CHP plants. The test run matrix is illustrated in Table 3.

2.2. Online corrosion probe The online corrosion probes (Fig. 2a) applied within the test runs consist of a sensor (Fig. 2b) which is placed on the top of a carrierlance. The sensor is temperature controlled and cooled by air allowing the simulation of heat exchanger tubes with different surface temperatures. It consists of 4 rings, which are made of a steel of choice, in this study 13CrMo4-5, P91 or 1.4541. When exposed to the flue gas, a layer of deposits and corrosion products forms on the surface of the sensor. This layer represents an electrolyte and allows the instantaneous determination of a corrosion signal with the three electrode rings which is proportional to the corrosion rate. The relation between the corrosion signal and the actual corrosion rate is determined with the mass loss of the mass loss ring. A more detailed description of the online corrosion probe can be found in [1,10–12]. Since the conductive layer starts to form as soon as the probe is exposed to the flue gas, the measured corrosion signal increases gradually from zero. As soon as a fully developed ionic layer has been formed on the surface, the corrosion signal is proportional to the corrosion rate. Therefore, in the initial phase of the measurement, which lasts about 300 h based on experiences with forest wood chips combustion, the actual corrosion rates cannot be correctly determined with the online corrosion probe. A detailed description of this measurement error is given in [1]. 2.3. Mass loss probe In order to gain data for correction of the measurement error from the online corrosion probe during the initial phase a special mass loss probe has been developed as described in [1]. The mass loss probe consists of an air-cooled carrier-lance with five test rings on top. The temperature of the test rings is controlled by cooling air which allows the simulation of a heat exchanger tube similar to the online corrosion probe. These rings have to consist of the same steel as the sensor of the online corrosion probe. For the measurement, the mass loss probe is exposed to the flue gas next to the online corrosion probe with a similar surface temperature set value. Individual test rings are removed from the probe after different times (24–340 h) to gain time-related mass losses. Hereby the trend of the corrosion rate during the initial phase of a test run can be determined. To determine the mass loss of the test rings, the rings are weighed before being exposed to the flue gas. After the test run, the corrosion products are removed according to ASTM G1-03 [13] and the mass loss is determined gravimetrically. Details regarding the mass loss probe are presented in [1].

2.1. Biomass test rig 2.4. Methodology for the evaluation of corrosion rates The test rig (Fig. 1) consists of a biomass grate furnace equipped with air staging and flue gas recirculation coupled with an electrically heated vertical tube (the so-called drop tube). The drop tube has a length of 3 m and an inner diameter of 0.15 m. The grate furnace can be operated between 12–50 kW fuel input power (related to the net calorific value) and the drop tube has an electrical input power of up to 60 kW. At the exit of the drop tube the measurement ports for the online corrosion probe and the mass loss probe are located. A more detailed description of the experimental facility is given in [1,9].

At the start of the test run the online corrosion probe and the mass loss probe are exposed to the flue gas and are operated at similar surface temperatures throughout the whole test run. The test run is divided into two phases: the initial phase and the variation phase. During the initial phase the operating conditions of the plant and the probes are kept constant. In this phase the trend of the corrosion rate is determined with the mass loss probe. When a fully developed ionic layer has been formed on the surface of the online corrosion probe the variation

Table 1 Fuel specifications according to EN 14961-1:2010.

Wood chips Waste wood

Origin

Particle

Moisture

Ash

Bulk density

1.1.4.4 1.1.4.5 1.3.1.1 1.3.2.1

P45Ac

M35 for 13CrMo4-5 M25 for P91 M25 for P91 M30 for 1.4541

A3.0

BD200

A10.0+ (21.8) for P91 A10.0+ (20.4) for 1.4541

BD250

P45Ac

S. Retschitzegger et al. / Fuel Processing Technology 142 (2016) 59–70

61

Table 2 Chemical compositions of 13CrMo4-5 (1.7335), P91 (1.4903) and 1.4541 in weight percent. Steel

C

Si

Mn

P

S

Al

Cu

Cr

Mo

Ni

Fe

13CrMo4-5 P91 1.4541

0.14 0.10 0.08

0.35 0.35 1.00

0.55 0.45 2.00

0.025 0.02 0.04

0.02 0.01 0.015

0.04 0.04 –

0.30 – –

0.93 8.75 18

0.50 0.95 –

0.30 0.40 10.5

Rest Rest Rest

phase is started. Here, flue gas temperatures and the probe surface temperatures are varied to determine their influence on the corrosion rate using the signal from the online corrosion probe. As described in detail in [1] the calibration factor for the calculation of corrosion rates from the corrosion signal is obtained with the mass loss occurring at the mass loss probe during the variation phase. The mass loss is combined with the integral over the measurement signal of the online corrosion probe during the variation phase. This calibration factor allows a correct calculation of the corrosion rates, since it does not take the wrong online corrosion probe data from the initial phase into account. This methodology is called “mass loss correction”. 2.5. Deposit probe Deposit probe measurements allow the investigation of deposition formation of ash particles on heat exchanger surfaces. The deposit probe applied consists of a temperature controlled air cooled carrierlance with a test ring on top. (Fig. 3). In order to determine the deposit formation only and to prohibit corrosion the test ring is made of the high-alloyed steel 1.4841. The probe is exposed to the flue gas for a certain period at a selected surface temperature. By gravimetric measurements of the test ring before and after the exposure to the flue gas the average deposit build-up rate can be calculated. This setup allows the simulation of heat exchanger tubes with different surface temperatures. The variation of the exposure time to the flue gas allows for the investigation of the time dependent deposit build-up. Furthermore, the chemical composition of the deposits can be determined using SEM/EDX (Scanning Electron Microscopy/Energy Dispersive X-ray Spectroscopy) analyses. 2.6. Fuel sampling and analyses The fuel sampling and subsequent analyses were performed as described in [1]. The fuel was sampled 4 times during the test run. The moisture content was determined according to the European Norm (EN) EN 14774-1 [14]. For further analyses, the samples were prepared according to EN 14780 [15]. The determination of the ash content was carried out according to EN 14775 [16]. The C, H, and N-contents were determined according to EN 15104 [17]. The determination of major and minor ash forming elements was carried out according to EN 15290 [18] and 15296 [19]. 2.7. SEM/EDX analyses SEM/EDX analyses were performed similar as presented in [5]. SEM/ EDX was used to determine the chemical composition of deposits from the deposit probe. For this purpose the deposit samples were coated with carbon before the analyses. The carbon coating ensures conductance of the surface and hence, prevents the samples from charging during the analyses. The analyses were performed using a Zeiss Gemini 982 (SEM-system) and a Noran Voyager (EDX-system). Table 3 Test run matrix.

Wood chips Waste wood

13CrMo4-5

P91

1.4541

x

x x

x

The electrodes of the online corrosion probes were analyzed subsequent to the test runs using SEM/EDX to investigate the structure and chemical composition of the deposits and corrosion products. For preparation, the electrode rings were embedded in resin and grinded with sand papers (silicon carbide) without using any fluid in order to avoid dissolution of sample species. Then the samples were also coated with carbon. The analyses were performed using a Zeiss Ultra 55 (SEMsystem) equipped with an EDAX Pegasus (EDX-system). 2.8. Test campaigns and operating conditions Within this paper the results of four test runs with different steels (13CrMo4-5, P91 and 1.4541) are presented. The biomass furnace was operated at constant load during all test runs. Each test run was divided into 2 phases: during the initial phase the probes were set to constant surface temperatures and the flue gas temperature at the measurement port was kept constant. During this time a fully developed ionic layer was formed on the surface of the corrosion probe. As soon as this layer had been formed, the variation phase got started. In this phase the flue gas temperature was varied in the range of 735 to 900 °C by electrical heating of the drop tube. For each flue gas temperature the probe surface temperatures were varied. These variations were performed between 400 and 560 °C for P91 and for 1.4541 and between 400 and 550 °C for 13CrMo4-5. For 13CrMo4-5 the surface temperature of 560 °C has not been investigated since the material is expected to start scaling at temperatures of 560 °C [20], 550 °C have therefore not been exceeded. Relevant operating data of the test runs are presented in Table 4. The operating conditions showed only small deviations. Therefore relevant impacts on the resulting corrosion rates were not to be expected. The fuel compositions and possible impacts on the corrosion rates are discussed in Section 3.1. 3. Results and discussion 3.1. Fuel analyses Fuel samples were taken in regular intervals from the fuel feeding system during both test runs with the results of the analyses given in Table 5. The analyses showed a typical composition of wood chips with bark with rather large standard deviations which had to be expected for a rather inhomogeneous fuel. The moisture content of the fuel used during the woodchips — 13CrMo4-5 test run was with 31.6 wt.% higher than for the wood chips — P91 test run with 20.5 wt.%. On the one hand this difference can influence the temperature in the primary combustion zone slightly and hereby, the release rates of ash forming elements to a minor extent. On the other hand, it resulted in different moisture contents in the flue gas of 15.4 vol.% wet basis for the 13CrMo4-5 test run and 12.3 vol.% wet basis for the P91 test run, both values at 8 vol.% O2 dry basis. Pujilaksono et al. [21] reported a 68% higher parabolic rate constant for the oxidation of iron in wet O2 containing 40 vol.% H2O compared to dry O2 at 500 °C. These results show that moisture in the flue gas can enhance the corrosion rate significantly. For the moisture contents typical for flue gases of biomass combustion system no studies are known to the authors. Since the moisture contents in the flue gas were comparable for both test runs it was assumed that the influence of moisture in the flue gas on the corrosion rate was also corresponding. The mean moisture contents of the waste wood were 24.0 wt.% for the P91 test run and 27.5 wt.% for the 1.4541

62

S. Retschitzegger et al. / Fuel Processing Technology 142 (2016) 59–70

Fig. 1. Scheme of the biomass grate furnace combined with a drop tube. Explanations: source: [1].

test run. Here, only small differences occurred between the two test runs. The mean ash contents of the wood chip fuels were almost the same with 2.7 and 2.8 wt.% for the two test runs. For waste wood the mean ash contents were 21.8 and 20.4 wt.%. Here a large difference to wood chips is given. Significant shares of soil (fine particles + sand) were found in the fuel. The contents of C, H and N were rather constant for all test runs. C and H were lower for waste wood compared to wood chips, mainly due to the high ash content. K, Na, S and Cl are considered as the most relevant elements for high temperature corrosion when firing chemically untreated woody biomass fuels. Their concentrations in the fuel remained rather constant over the wood chips test runs. Besides K, Na, S and Cl also the heavy metals Zn and Pb are usually relevant for high temperature corrosion when firing waste wood [22,23]. The concentrations of these elements were rather constant over the test runs. The heavy metal concentrations found in the waste wood significantly exceeded the ones from wood chips as expected, since the waste wood contained chemically treated components (e.g. Zn: 165 and 195 mg/kg d.b. for waste compared to 15 and 18 mg/kg d.b. for wood chips). This resulted in a higher potential for corrosion for waste wood. The S content in the fuel during the wood chips — P91 test run was higher at an almost similar Cl content resulting in a higher molar ratio of 2S/Cl of 9.3 whereas for wood chips — 13CrMo4-5 it was 7.6. This indicated a lower risk for Cl-induced active oxidation for wood chips — P91 [24]. Mean 2S/Cl ratios for waste wood were calculated with 1.7 (P91) and 2.2 (1.4541). Within the waste wood test runs a good

comparability was given regarding the risk for Cl-induced active oxidation. Compared to wood chips, it was significantly higher. Generally, it can be stated, that the fuel compositions mainly deviated within their standard deviation, therefore, significant impacts on the results were not likely. The potential for high temperature corrosion is generally clearly higher for waste wood compared to wood chips due to the elevated heavy metal concentrations and the lower 2S/Cl values. 3.2. Deposit formation In order to characterize and compare the deposit formation processes occurring within the test runs, deposit probe measurements were performed for sampling times of 2 h with surface temperatures of the deposit probes of 480 °C and 560 °C. These measurements were done at flue gas temperatures of 745 °C for the wood chips — 13CrMo4-5 test run, at 760 °C for the wood chips — P91 test run and at 740 °C (P91) respectively 735 °C (1.4541) for the waste wood test runs. The resulting rates of deposit built up (RBU) are presented in Table 6. For the test run with wood chips — P91 the RBUs were higher than for the test run with wood chips — 13CrMo4-5 which most likely was a result of the slightly higher K content of the fuel while the Si content was lower. This is likely to result in a higher release of K during combustion which increases the concentration of ash forming elements in the flue gas [24–26]. While ash chemistry is more complex than just the interaction between K and Si, the consideration of these two elements allows an estimation of the K release [24]. The measured deposit rates indicate that a faster deposit build-up on the online corrosion probe and the mass loss probe during the initial

Fig. 2. Scheme of the online corrosion probe. Explanations: a) corrosions probe; b) sensor with three electrodes (1–3) and the mass loss ring (M); source [13].

S. Retschitzegger et al. / Fuel Processing Technology 142 (2016) 59–70

63

temperatures might result from higher condensation rates but also the inhomogeneities of the fuel, especially variations in the ash content, have to be considered to have caused differences in the results. The chemical compositions of the deposits have been analyzed at three positions (windward, windward + 50°, leeward) by means of SEM/EDX as described in [5]. The results of the analyses are presented in Fig. 4. The deposits resulting from wood chip combustion mainly consisted of K and S, their ratio indicating the presence of K2SO4. On the windward side increased amounts of Ca and Si as well as small fractions of Mg were found. These elements typically form coarse fly ashes and are deposited mainly by inertial impaction [27]. During the 13CrMo4-5 test run higher contents of Si occurred in the coarse fly ash which resulted from higher concentrations in the fuel. Na was found in small concentrations compared to K which had to be expected due to the significantly lower content of Na in the fuel. Cl could only be found in small amounts in the case of 480 °C probe surface temperature during the P91 test run. This indicates that the condensation temperature of the gaseous chlorides was between 480 °C and 550 °C. For waste wood combustion the deposits on the windward side were to a certain extent comparable to the ones from wood chips. Here, the deposits consisted to large amounts of Ca, Mg and Si from

Fig. 3. Scheme of the deposit probe. Explanations: TIC — temperature measurement and control. Explanations: source: [5].

phase of the test run had to be expected for the P91 test run. Since the differences in deposit formation result from the different fuel compositions, the use of two different batches of wood chips has been to some extent negatively influencing the comparability of the test runs but the uncertainties are considered to be within acceptable ranges. For waste wood the deposition rates were considerably higher with RBUs ranging from 14.1 to 19.2 g/m2 ∗h. The higher RBUs were a direct result from the significantly higher ash content and the high concentrations of volatile ash forming elements of the waste wood compared to wood chips (see Section 3.1). For waste wood the RBUs ranged from 14.1 to 19.2 g/m2 ∗ h with higher RBUs at the lower probe surface temperature. On the one hand the higher RBUs at lower surfaces

Table 4 Relevant operating conditions of the test rig and the two probes. Explanations: d.b. — dry basis; temperatures during the variation phase are targeted values.

Parameter

Fuel power input O2 CO2 CO Flue gas temperature at measurement port Flue gas velocity at entrance of measurement port Probe surface temperature

Unit

kW vol.% d.b. vol.% d.b. mg/m3 at 0 °C and 101 325 Pa, d.b. °C m/s °C

Wood chips 13CrMo4-5

Wood chips P91

Waste wood P91

Waste wood 1.4541

Mean value ± standard deviation

Mean value ± standard deviation

Mean value ± standard deviation

Mean value ± standard deviation

Initial phase

Initial phase

Initial phase

Initial phase

Variation phase

23.2 ± 2.1 8.6 ± 1.6 12.0 ± 1.6 3.1 ± 2.1

Variation phase

23.0 ± 2.3 8.3 ± 1.6 12.4 ± 1.6 4.0 ± 2.1

Variation phase

23.4 ± 2.0 9.7 ± 1.2 11.0 ± 1.2 9.5 ± 6.9

Variation phase

23.3 ± 2.8 9.6 ± 1.6 11.2 ± 1.7 8.5 ± 6.3

745 ± 12.9

745/815/860

764 ± 32.9

740/810/860

741 ± 20.0

750/840/900

728 ± 16.3

735/835/865

2.7 ± 0.1

2.7/2.9/3.0

2.9 ± 0.1

2.9/3.0/3.1

2.6 ± 0.7

2.6/2.7/2.8

2.7 ± 0.7

3.0/3.2/3.2

480 ± 0.1

400/480/550

480 ± 0.1

400/480/560

480 ± 0.3

400/480/550

480 ± 0.2

400/480/560

Table 5 Results of fuel analyses. Explanations: w.b. — wet basis; d.b. — dry basis; Std.-dev. — standard deviation.

Moisture content Ash content (550 °C) C H N S Cl Ca Si Mg Al Na K Fe P Mn Zn Pb 2S/Cl

wt.% w.b. wt.% d.b. wt.% d.b. wt.% d.b. wt.% d.b. mg/kg d.b. mg/kg d.b. mg/kg d.b. mg/kg d.b. mg/kg d.b. mg/kg d.b. mg/kg d.b. mg/kg d.b. mg/kg d.b. mg/kg d.b. mg/kg d.b. mg/kg d.b. mg/kg d.b. mol/mol

Wood chips — 13CrMo4-5

Wood chips — P91

Waste wood — P91

Waste wood — 1.4541

n = 4 samples

n = 4 samples

n = 5 samples

n = 4 samples

Mean

Std.-dev

Mean

Std.-dev

Mean

Std.-dev

Mean

Std.-dev

31.6 2.7 48.1 6.0 0.3 262 117 4723 3654 614 615 128 1875 313 242 88 15 4.1 7.6

3.9 1.2 0.4 0.1 0.1 70 71 1207 3009 193 424 92 449 237 71 26 5 4.6 2.7

20.5 2.8 49.6 6.0 0.5 400 95 4795 2938 616 671 92 2205 390 428 603 28 8.8 9.3

1.2 0.7 0.8 0.0 0.2 124 57 1071 1553 136 245 31 669 125 157 446 12 0.6 2.3

24.0 21.8 42.1 5.1 1.7 1646 2494 41,280 30,940 4326 8300 2708 4450 7660 1102 235 165 38 1.7

4.1 7.1 – – – 227 1316 10,756 11,234 960 2648 598 774 5907 291 87 59 19 0.6

27.5 20.4 42.8 5.2 1.7 1617 1463 38,533 27,700 4047 7430 2277 4250 5043 1129 208 195 15 2.2

4.4 6.8 – – – 407 144 13,158 6975 1347 2010 551 747 1707 383 44 95 10 0.5

64

S. Retschitzegger et al. / Fuel Processing Technology 142 (2016) 59–70

Table 6 Results of deposit probe measurements Explanations: RBU — rate of deposit build up; sampling time: 2 h. Surface temperature [°C]

RBU [g/m2 ∗h] Wood chips — 13CrMo4-5

Wood chips — P91

Waste wood — P91

Waste wood — 1.4541

480 560

1.9 1.8

2.7 2.7

19.2 14.5

14.3 14.1

coarse fly ashes but also K, S and Cl were found. Significant differences from the results for wood chips were found at windward +50° and at the leeward side. For waste wood the deposits mainly consisted of K, Na and Cl, indicating the presence of KCl and NaCl but also shares of S were found, indicating K- and Na-sulfates. Another difference was an obvious share of Zn (up to 2.0 mol%) in the waste wood deposits. No significant deviations regarding the deposit composition have been found between the two test runs for each fuel (wood chips respectively waste wood). Therefore, the deposit analyses formed a good basis to study their effect on the different steels. However, since corrosion is the reaction of the surroundings (deposits, flue gas, …) with the steel, certain deposit compositions do not necessarily result in the same corrosion processes for different steels. 3.3. Corrosion rates Fig. 5 shows the measured corrosion (material loss determined with the mass loss probe) in [mm] for the two materials investigated for wood chips combustion. During the initial phase the corrosion followed a paralinear trend which corresponds with data from literature [28,29]. For wood chips — 13CrMo4-5 a corrosion of 0.045 mm was found after 312 h, whereas for wood chips — P91 the corrosion was 0.028 mm after 309 h. These results showed a generally better resistance of P91 towards high temperature corrosion during the initial phase. For waste wood – P91 a corrosion of 0.073 mm was found after the initial phase whereas for waste wood – 1.4541 this value resulted to 0.036 mm (Fig. 6). As for the wood chips test runs, the higher alloyed steel (in this case 1.4541) showed a generally better corrosion resistance. For the waste wood test runs the paralinear fits generally

describes the corrosion trend in the initial phase but are less accurate than for the wood chips test runs. A possible reason for the stronger deviations is the removal of the corrosion products (see Section 2.3). The corrosion products of different steels behave differently, e.g. they may adhere to the not corroded material firmly at certain positions or they may be rather loose. For the removal of the corrosion products of P91 and 1.4541 it was more difficult to only remove corrosion products without dissolving the not corroded material. Therefore, the scattering of the measured corrosion values is more distinctive for these materials compared to 13CrMo4-5. The mass loss correction of the corrosion rates was based on the mass loss (Δ g) in [mm] obtained during the variation phase. These values were coupled with the data obtained from the online corrosion probe during the variation phase, as described in detail in [1]. The original corrosion rates and the mass loss corrected corrosion rates as well as the flue gas temperatures and the probe surface temperatures during the variation phases for the wood chips test runs are plotted in Fig. 7. For wood chips — 13CrMo4-5 the mass loss correction reduced the measured corrosion rates by 55% whereas for wood chips — P91 the reduction amounted to 32%. Hence the measurement error (described in detail in [1]) resulting from the initial phase, was considerably higher for wood chips — 13CrMo4-5. However, it should be highlighted that these values may depend on the individual test run conditions. Therefore, they cannot be applied generally and additional test runs for confirmation should be performed in future. The same applies for the mass loss correction obtained for the waste wood test runs. The original corrosion rates and the mass loss corrected corrosion rates as well as the flue gas temperatures and the probe surface temperatures during the variation phases for the wood chips test runs are plotted in Fig. 8. The data plotted in Figs. 7 and 8 show, that the corrosion rates immediately followed the variations of the flue gas temperature and the probe surface temperature for all test runs. This outlines the applicability of the online corrosion probe for different fuels and different heat exchanger materials. For waste wood — P91 the mass loss correction reduced the measured corrosion rates by 55% whereas for waste — 1.4541 the reduction amounted to 77%. Hence the measurement error resulting from the initial phase, is higher for the higher alloyed steel. The differences in

Fig. 4. Results of SEM/EDX analyses of the deposits sampled at different positions on the deposit probes. Explanations: sampling time: 2 h.

S. Retschitzegger et al. / Fuel Processing Technology 142 (2016) 59–70

65

Fig. 5. Mass losses determined with the mass loss probe as well paralinear fits of the mass losses during the initial phase of the wood chips test runs. Explanations: data for 13CrMo4-5 adapted from: [1].

Fig. 6. Mass losses determined with the mass loss probe as well paralinear fits of the mass losses during the initial phase of the waste wood test runs.

the reduction of corrosion rates show that a comparison of different short-term test runs using online corrosion probes is inaccurate without mass loss correction. The corrosion rates for certain probe surface temperatures are plotted against the flue gas temperature for the wood chips test runs in Fig. 9 and for the waste wood test runs in Fig. 10. In these Figures only certain flue gas and probe surface temperatures are shown for comparison reasons. A trend of increasing corrosion rates with increasing flue gas temperatures can be seen. Furthermore, it can

be seen that increasing probe surface temperatures resulted in increasing corrosion rates too. These general trends were similar for all test runs. For the wood chips test runs below 750 °C flue gas temperature the corrosion rates determined were comparable for 13CrMo4-5 and P91. Corrosion rates of 0.006 μm/h were determined at a probe surface temperature of 400 °C. At the highest probe surface temperatures investigated the rates were 0.16 μm/h for 13CrMo4-5 (at 550 °C) respectively 0.18 μm/h for P91 (at 560 °C). For flue gas temperatures exceeding 800 °C

Fig. 7. Corrosion rates with and without mass loss correction for the wood chips test runs in dependence of probe surface and flue gas temperature. Explanations: data for 13CrMo4-5 adapted from: [1].

66

S. Retschitzegger et al. / Fuel Processing Technology 142 (2016) 59–70

Fig. 8. Corrosion rates with and without mass loss correction for the waste wood test runs in dependence of probe surface and flue gas temperature.

the corrosion rates of P91 only slightly increased at a probe surface temperature of 400 °C. For 13CrMo4-5 however the corrosion rates showed a clear dependence on the flue gas temperatures with values up to 0.054 μm/h at 860 °C. A similar trend was found for the highest probe surface temperatures. For 13CrMo4-5 a corrosion rate of 0.34 μm/h (at 550 °C) and for P91 a rate of 0.30 μm/h (at 560 °C) was determined. These results show, that P91 has a higher resistance towards high temperature corrosion than 13CrMo4-5 at high flue gas temperatures exceeding 800 °C but no significant difference is given at lower flue gas temperatures in the range of 740 °C. Although the moisture content in the flue gas for the 13CrMo4-5 test runs was higher than for P91 test runs (15.4 compared to 12.3 vol.%) this small difference is not considered to be responsible for the higher corrosion rates found for 13CrMo4-5. In case of the waste wood test runs (Fig. 10), significantly lower corrosion rates were determined for 1.4541 compared to P91. The highest corrosion rate for 1.4541 was 0.041 μm/h (at the probe surface temperature TS = 560 °C and flue gas temperature TFG = 860 °C) which was still beneath the lowest corrosion rate for P91 with 0.049 μm/h (at TS = 400 °C and TFG = 750 °C). This shows that 1.4541 has a significantly higher resistance against high temperature corrosion compared to P91. The corrosion rates for P91 for wood chips combustion and waste wood combustion were generally in the same range, but showed different distinctive dependencies on the influencing temperatures. The corrosion rates obtained for waste wood were lower than the ones for wood chips

combustion at low probe surface temperatures (TS = 400 °C) but were higher at high probe surface temperatures (TS = 560 °C). For TS = 400 °C and TFG = 740 °C (wood chips) respectively TFG = 750 °C (waste wood) the corrosion rates determined were 0.006 μm/h for wood chips and 0.049 μm/h for waste wood. For TS = 560 °C the corrosion rates were 0.18 μm/h for wood chips and 0.15 μm/h for waste wood. This shows, that the influence of the probe surface temperature was significantly higher for wood chips combustion. A similar trend was found for the higher flue gas temperatures. Here again, the corrosion rates for waste wood were lower than the ones for wood chips at TS = 400 °C but were higher at TS = 560 °C. The reasons for these differences have not been identified yet and are subject of further investigations. One possible reason is a difference in the prevalent corrosion mechanism. For waste wood combustion besides significant amounts of Cl in deposits also Zn was found in the deposits (see Section 3.2). This may cause a difference in corrosion kinetics. However, this topic has not been further investigated in this study. Another possible reason is an error within the methodology for the determination of the mass loss with the mass loss probe. The method has been developed using wood chips and 13CrMo4-5 as steel [1]. As presented in Fig. 6 the mass losses determined for P91 and 1.4541 scatter significantly stronger than for 13CrMo4-5 (Fig. 5). This means, also the last mass loss, which is relevant for the determination of Δg may contain a certain error which directly influences the calculation of the corrosion rates. Therefore, it is suggested to further develop the methodology and use a second mass loss ring for the determination of the last mass loss to ensure a higher reliability of the results.

Fig. 9. Corrosion rates in dependence of the flue gas temperature for certain probe surface temperatures. Explanations: corrosion rates are mean values at certain flue gas temperatures (±10 °C) and certain corrosion probe temperatures (±1 °C); error bars show the standard deviations; data for 13CrMo4-5 adapted from: [1].

S. Retschitzegger et al. / Fuel Processing Technology 142 (2016) 59–70

67

Fig. 10. Corrosion rates in dependence of the flue gas temperature for certain probe surface temperatures. Explanations: corrosion rates are mean values at certain flue gas temperatures (±10 °C) and certain corrosion probe temperatures (±1 °C); error bars show the standard deviations.

3.4. Analyses of corrosion products and deposits In order to determine the chemical composition as well as the structure of the corrosion products and the deposit layer on the online corrosion probe, SEM/EDX element mappings have been performed at the windward side and the leeward side of the probes. Since the general information obtained for the windward side and the leeward side was the same for the wood chips test runs as well as for the waste wood test runs, only the mappings of the windward side are presented in Figs. 11 and 12. The absolute thickness of corrosion layers presented in these figures might deviate from the resulting corrosion rates. On the one hand, the layer thickness was not constant over the whole ring but this did not affect the determination of corrosion rates, since the whole corroded material is removed. On the other hand, the measurement times and temperatures were different for all test runs resulting in different amounts of corrosion. The deposit layers mainly consisted of K, Si, Ca and S as well as O (see also spot analyses in Figs. 13 and 14). These results indicated the presence of K- and Ca-sulfates as well as Ca- and Si-oxides which is in good agreement with the results from the deposit probes. Solid K- and

Ca-sulfates as well as the oxides found usually have only a minor effect on the corrosion behavior [30]. Therefore, no significant impact on the corrosion process has been concluded from these deposits. For waste wood the shape of the deposits indicated, that sintering respectively melting of deposits has occurred during the test runs. Melting deposits when firing waste wood have already been reported by other authors [31,32] and are mainly referred to the combination of different alkali metal and heavy metal sulfates and chlorides leading to low eutectic melting temperatures. Zn has been found in the deposits of waste wood — 1.4541 (see Fig. 14). For waste wood also large amounts of Cl were found during deposits probe measurements (see Section 3.2) but Cl has not been found in the deposits on the corrosion rings (neither on the windward nor on the leeward side). The corrosion layers were of special interest since different steels have been used. These layers differed significantly between the individual test runs. In case of wood chips — 13CrMo4-5 one compact layer consisting of Fe and O as well as small concentrations of Cr was found. For wood chips — P91 the corrosion layer consisted of different sub-layers. On the outer side (close to the flue gas) no clear boundary between deposits and corrosion products was found. K,

Fig. 11. SEM/EDX element mappings of the corrosion probe rings on the windward side — wood chips test runs.

68

S. Retschitzegger et al. / Fuel Processing Technology 142 (2016) 59–70

Fig. 12. SEM/EDX element mappings of the corrosion probe rings on the windward side — waste wood test runs.

Ca and S were mixed with Fe-oxides here. Towards the corrosion front, this mixed layer was alternately followed by a Cr-rich and a Fe-rich layer. Similar results showing separated iron oxide and chromium oxide layers were also reported in [7,22]. The separation most likely resulted from the higher affinity of Cr to O compared with Fe to O. Therefore, a Cr-oxide layer was formed on the outside (flue gas side), followed by a Fe-oxide layer. The alteration was assumed to be the result of spallation of the corrosion layer. The spallation most likely happened due to the variations of the probe surface temperature during the variation phase of the test run. After each spallation the Cr-oxide formation was favored again resulting in multiple Cr-rich and Fe-rich layers. A similar separation of a Cr-rich and a Fe-rich layer was found for waste wood — P91. The formation mechanism here was assumed to be similar to wood chips — P91. In the case of waste wood — P91 the corrosion layer was also spalled. On the inside of the corrosion layer, an enrichment of S was found. Spot analyses of this S-rich layer showed almost only Fe and S indicating this layer to consist of FeS or FeS2 (Spot 5 in Fig. 13). The corrosion layer of waste wood — 1.4541 was a mix of Fe, Cr and Ni-Oxides as well as Na, K and S. The spot analyses (Fig. 14) confirmed the result of the element mappings that here the corrosion layer was mixed with depositions. This was most likely again a result of the variations of the probe surface temperature. Here also a S-enrichment in the absence of O was found (Spot 5 in Fig. 14).

For the wood chips test run two findings are of special interest at the corrosion front. First an enrichment of S was found in case of 13CrMo4-5. This was also found by Retschitzegger et al. [5] as well as by Gruber et al. [6] under comparable test run conditions at the same test rig. Gruber et al. concluded that the S enrichment consisted of iron sulfides which were embedded in a matrix of iron oxides. The formation of iron sulfides have already been studied in the past (e.g. Gesmundo et al. [33] and Gilewicz-Wolter [34]) and can most likely be explained by the reaction of gaseous SO2 with the steel or the iron oxide in this case [30]. Secondly, almost no Cl was found at the corrosion front in both cases. Only in case of 13CrMo4-5 one spot of Cl was seen, which correlated with K and indicated the presence of KCl. Cl at the corrosion front is considered to be necessary for Clinduced corrosion. However, Cl-induced corrosion may be relevant based on the deposit probe investigations since Cl has been found in deposits sampled at 480 °C probe surface temperature. Due to the temperature variations performed with the online corrosion probe two effects may have been caused. At high probe surface temperatures already deposited chlorides and active Cl may have been evaporated. On the other hand at times of low probe surface temperatures increased condensation of chlorides may have occurred. This Cl could have been active for some time until the temperature would be increased again and Cl would have disappeared. Based on these results Cl-induced corrosion cannot be excluded as a corrosion mechanism in addition to oxidation and sulfidation of the steel.

Fig. 13. Spot and area analyses for waste wood — P91.

S. Retschitzegger et al. / Fuel Processing Technology 142 (2016) 59–70

69

Fig. 14. Spot and area analyses for waste wood — 1.4541.

For the waste wood test runs Cl was found at the corrosion front, which indicated Cl-induced corrosion to be a relevant corrosion process. In case of waste wood — P91 a severe attack of the ground material is visible in Fig. 13. Small area analyses of these areas showed a mixture of Cl, Fe, Cr and O (Fig. 13 — analyses 2 and 3). However, such an extreme corrosive attack was considered unlikely to happen in the short time of the test run. Metal chlorides are known to be highly hygroscopic and may therefore alter the sample after preparation. Salmenoja [35] reported hygroscopic interfaces between the base metal and the corroded scale supporting this assumption. Therefore, it was concluded that this corrosive attack was an artifact from sample preparation and analysis which occurred after the sample had been polished. The analyses of the samples from waste wood — 1.4541 (Fig. 14) showed that Cl-rich areas were separated from the O-rich areas. This finding supports the Cl-induced corrosion mechanism suggesting an attack of the ground material by Cl which forms Fe-chlorides that are later on oxidized. Although this separation existed, a certain mixing of Cl and O still maintained, as presented in Fig. 14, analyses 6 and 7. 4. Summary and conclusions The high-temperature corrosion behavior of the boiler steels 13CrMo4-5, P91 and 1.4541 has been evaluated during short-term test runs, using a biomass-fired grate furnace combined with an electrically heated drop tube. For the test runs forest wood chips (13CrMo4-5 and P91 as boilers steels) and waste wood (P91 and 1.4541) have been used as fuels. Online corrosion probes and a mass loss probe have been used applying a methodology developed in a previous study [1] to correct for a measurement error occurring during short-term measurements with online corrosion probes. This methodology is called mass loss correction. For wood chips — 13CrMo4-5 the mass loss correction reduced the measured corrosion rates by 55% whereas for wood chips — P91 the reduction amounted to only 32%. The reduction of corrosion rates for waste wood – P91 amounted to 55% and to 77% for waste wood – 1.4541. For the waste wood test runs there was higher scatter in the mass loss data compared to the wood chips test runs. This resulted in less accurate paralinear fits and thereby caused higher error margins for the waste wood test runs. However, the results still illustrate, that a comparison of different short-term test runs using online corrosion probes contains significant errors without a mass loss correction. Corrosion rates were determined for steel surface temperatures between 400 and 550 °C for 13CrMo4-5 and between 400 and 560 °C for P91 and 1.4541. Flue gas temperatures ranging from 740 to 900 °C have been investigated. In the wood chips test runs at flue gas temperatures below 800 °C basically no difference occurred for all probe surface temperatures investigated between the 13CrMo4-5 and P91. For flue gas temperatures exceeding 800 °C P91 showed lower corrosion rates than 13CrMo4-5. For waste wood combustion, 1.4541 resulted in significantly lower corrosion rates than P91 for all the investigated probe surface and flue gas temperatures. P91 has been investigated for both fuels. In comparison,

the corrosion rates for waste wood combustion were higher at low probe surface temperatures (400 °C). However, at higher probe surface temperatures the corrosion rates determined for wood chips combustion exceeded the ones for waste wood combustion clearly. The reasons for these differences have not been identified yet and are subject of further investigations. For wood chips fired CHP plants utilizing high flue gas temperatures above 800 °C at superheater inlet P91 seems to be a more suitable material than 13CrMo4-5. For more corrosive environments 1.4541 can be recommended due to its higher corrosion resistance. SEM/EDX analyses of the deposits sampled with a deposit probe showed that Cl-induced corrosion may play a role in the corrosion mechanism for wood chips combustion. Cl has been found in the deposits sampled at probe surface temperatures of 480 °C but at higher probe surface temperatures no Cl could be identified. Therefore, Clinduced corrosion cannot be excluded as a corrosion mechanism in addition to oxidation and sulfidation of the steel. For waste wood combustion significant concentrations of Cl were found in the deposits. Therefore, Cl-induced corrosion is highly relevant. Based on SEM/EDX analyses of the corrosion probe rings, no Cl could be found in the deposits or at the corrosion front for the wood chips test runs. Since temperature variations were performed with the online corrosion probe, Cl may have been evaporated at the higher temperatures. For the waste wood test runs Cl was mainly found at the corrosion front. This confirms the mechanism of Cl-induced corrosion. But also here, no Cl was determined in the deposits. Therefore, in order to evaluate the chemistry of deposits and hereby corrosion mechanisms in detail, experiments at constant flue gas and probe surface temperature should be performed as temperature variations may change the chemistry of deposits as well their structure.

Acknowledgments This article is the result of a project carried out within the framework of the Austrian COMET program, which is funded by the Republic of Austria as well as the federal provinces of Styria, Lower Austria and Burgenland. Within the COMET program the work was also funded by the companies BIOS BIOENERGIESYSTEME GmbH, EnBW Energie Baden-Württemberg AG, HET Heiz- & Energietechnik EntwicklungsGmbH, Josef Bertsch GmbH & Co and Viessmann Holzfeuerungsanlagen GmbH (Grant number: 834018).

References [1] S. Retschitzegger, T. Gruber, T. Brunner, I. Obernberger, Short term online corrosion measurements in biomass fired boilers. Part 1: application of a newly developed mass loss probe, Fuel Process. Technol. 137 (2015) 148–156. [2] Y.S. Li, M. Spiegel, S. Shimada, Corrosion behaviour of various model alloys with NaCl KCl coating, Mater. Chem. Phys. 93 (2005) 217–223. [3] M.A. Uusitalo, P.M.J. Vuoristo, T.A. Mäntylä, High-temperature corrosion of coatings and boiler steels below chlorine-containing salt deposits, Corros. Sci. 46 (2004) 1311–1331.

70

S. Retschitzegger et al. / Fuel Processing Technology 142 (2016) 59–70

[4] H.J. Grabke, E. Reese, M. Spiegel, The effects of chlorides, hydrogen chloride, and sulfur dioxide in the oxidation of steels below deposits, Corros. Sci. 37 (1995) 1023–1043. [5] S. Retschitzegger, T. Brunner, I. Obernberger, B. Waldmann, Assessment of online corrosion measurements in combination with fuel analyses and aerosol and deposit measurements in a biomass combined heat and power plant, Energy Fuel 27 (2013) 5670–5683. [6] T. Gruber, K. Schulze, R. Scharler, I. Obernberger, Investigation of the corrosion behaviour of 13CrMo4–5 for biomass fired boilers with coupled online corrosion and deposit probe measurements, Fuel 144 (2015) 15–24. [7] J. Lehmusto, P. Yrjas, B.J. Skrifvars, M. Hupa, High temperature corrosion of superheater steels by KCl and K2CO3 under dry and wet conditions, Fuel Process. Technol. 104 (2012) 253–264. [8] EN 14961-1:2010 Solid biofuels – Fuel specifications and classes – Part 1: General requirements; German version EN 14961-1:2010, March 2010 (Published). [9] T. Gruber, R. Scharler, I. Obernberger, Application of an empirical model in CFD simulations to predict the local high temperature corrosion potential in biomass fired boilers, Biomass Bioenergy (2015)http://dx.doi.org/10.1016/j.biombioe.2015.02. 024. [10] B. Waldmann, Korrosion in Anlagen zur Thermischen Abfallverwertung: Elektrochemische Korrosionserfassung und ModellbildungPhD–Thesis University of Augsburg, Germany, 2007. [11] S. Maisch, Identifikation und Quantifizierung von Korrosionsrelevanten Parametern in Müllverbrennungsanlagen Mittels Charakterisierung der Deponierten Partikel und Elektrochemischer Online-MessungenPhD–Thesis University of Augsburg, Germany, 2011. [12] C. Deuerling, B. Waldmann, Innovative Messtechnik zur kombinierten Korrosions- und Materialforschung im Kraftwerksbereich, 43, Kraftwerkstechnisches Kolloquium2011 2011 (Dresden). [13] ASTM G1 – 03, Standard Practice for Preparing, Cleaning, and Evaluating Corrosion Test Specimens, 2011. [14] EN 14774-1:2009 Solid Biofuels. Determination of Moisture Content. Oven dry method. Total moisture. Reference method, January 2010 (Published). [15] EN 14780:2011 Solid Biofuels. Sample Preparation, June 2011 (Published). [16] EN 14775:2009 Solid Biofuels. Determination of Ash Content, March 2010 (Published). [17] EN 15104:2011 Solid Biofuels. Determination of Total Content of Carbon, Hydrogen and Nitrogen. Instrumental Methods, February 2011 (Published). [18] EN 15290:2011 Solid Biofuels. Determination of Major Elements. Al, Ca, Fe, Mg, P, K, Si, Na and Ti, February 2011 (Published). [19] EN 15296:2011 Solid Biofuels. Conversion of Analytical Results from One Basis to Another, February 2011 (Published).

[20] T.J.A. Richardson, Shreir's Corrosion, 1st edition Elsevier Science, Amsterdam, Netherlands, 2009, ISBN 978-0444527882. [21] B. Pujilaksono, T. Jonsson, M. Halvarsson, J.E. Svensson, L.G. Johansson, Oxidation of iron at 400–600 °C in dry and wet O2, Corros. Sci. 52 (2010) 1560–1569. [22] D. Bankiewicz, S. Enestam, P. Yrjas, M. Hupa, Experimental studies of Zn and Pb induced high temperature corrosion of two commercial boiler steels, Fuel Process. Technol. 105 (2012) 89–97. [23] F. Frandsen, Utilizing biomass and waste for power production—a decade of contributing to the understanding, interpretation and analysis of deposits and corrosion products, Fuel 84 (2005) 1277–1294. [24] P. Sommersacher, T. Brunner, I. Obernberger, Fuel indexes: a novel method for the evaluation of relevant combustion properties of new biomass fuels, Energy Fuel 26 (2012) 380–390. [25] J.N. Knudsen, Volatilization of Inorganic Matter During Combustion of Annual BiomassPhD–Thesis Technical University of Denmark, Lyngby, Denmark, 2003, ISBN 87-91435-11-0. [26] S.C. van Lith, Release of Inorganic Elements During Wood-firing on a GratePhD–Thesis Technical University of Denmark, Lyngby, Denmark, 2005, ISBN 87-91435-29-3. [27] T. Lauren, Methods and Instruments for Characterizing Deposit Buildup on Heat Exchangers in Combustion PlantsLicentiate Thesis Abo Akademi, Abo/Turku, Finland, 2008. [28] P.J. Gellings, Introduction to Corrosion Prevention and Control, Delft University Press, Delft, Netherlands, 1985, ISBN 9062751725. [29] P. Kofstad, High-temperature Corrosion, Elsevier Applied Science Publishers LTD., London, England, 1988, ISBN 1851661549. [30] R.A. Antunes, M.C.L. de Oliveira, Corrosion in biomass combustion: a materials selection analysis and its interaction with corrosion mechanisms and mitigation strategies, Corros. Sci. 76 (2013) 6–26. [31] S. van Lith, C. Buchardt, F.J. Frandsen, Final Report (D20) for SES6-CT-2003–502679: Ash and Aerosol Related Problems in Biomass Combustion and Co-firing (BIOASH), Task 2.3: Lab-Scale Investigation of the Potential for Corrosion of Deposits, CHEC Research Centre, Technical University of Denmark, Denmark, 2007. [32] M. Spiegel, Salt melt induced corrosion of metallic materials in waste incineration plants, Mater. Corros. 50 (1999) 373–393. [33] F. Gesmundo, D.J. Young, S.K. Roy, The high temperature corrosion of metals in sulfidizing–oxidizing environments: a critical review, High Temp. Mater. Process. 8 (1989) 149–189. [34] J. Gilewicz-Wolter, Study of iron oxdiation in sulfur dioxide atmospheres by means of the 35S radioistiope, Oxid. Met. 11 (1977) 81–91. [35] K. Salmenoja, Field and Laboratory Studies on Chlorine-induced Superheater Corrosion in Boilers Fired with BiofuelsPhD–Thesis Abo Akademi, Abo/Turku, Finland, 2000.